The Web services composition system with static configuration lacks of adaptive capacity to both the QoS variability of component services and the failure-prone environment. The static configuration can not reflect the dynamic evolution process of Web services composition when the number of component services increases continually. A method of QoS-aware and dynamic configuration for Web services composition is presented. The functional dependency relationship of Web services is described formally via an extended Petri net to form a configuration scheme. The configuration scheme is verified by two algorithms

which are used to confirm there are no loop dependency relationship and illegal end state in the configuration scheme. An optimal configuration selection algorithm is used to search the configuration with best QoS from the configuration scheme. Simulation experiments are conducted to compare the effects on the satisfaction rate of user request for service among the dynamic

static and random configurations.The results show that the QoS of composition service can be improved greatly by using the proposed modeling method and algorithms.